The Pathological Classification of Prostate Lesions in Genetically Engineered Mice (GEM) is the result of a directive from the National Cancer Institute Mouse Models of Human Cancer Consortium Prostate Steering Committee to provide a hierarchical taxonomy of disorders of the mouse prostate to facilitate classification of existing and newly created mouse models and the translation to human prostate pathology. The proposed Bar Harbor Classification system is the culmination of three meetings and workshops attended by various members of the Prostate Pathology Committee of the Mouse Models of Human Cancer Consortium. A 2-day Pathology Workshop was held at The Jackson Laboratory in Bar Harbor, Maine, in October 2001, in which study sets of 93 slides from 22 GEM models were provided to individual panel members. The comparison of mouse and human prostate anatomy and disease demonstrates significant differences and considerable similarities that bear on the interpretation of the origin and natural history of their diseases. The recommended classification of mouse prostate pathology is hierarchical, and includes developmental, inflammatory, benign proliferative, and neoplastic disorders. Among the neoplastic disorders, preinvasive, microinvasive, and poorly differentiated neoplasms received the most attention. Specific criteria were recommended and will be discussed. Transitions between neoplastic states were of particular concern. Preinvasive neoplasias of the mouse prostate were recognized as focal, atypical, and progressive lesions. These lesions were designated as mouse prostatic intraepithelial neoplasia (mPIN). Some atypical lesions were identified in mouse models without evidence of progression to malignancy. The panel recommended that mPIN lesions not be given histological grades, but that mPIN be further classified as to the absence or presence of documented associated progression to invasive carcinoma. Criteria for recognizing microinvasion, for classification of invasive gland-forming adenocarcinomas, and for characterizing poorly differentiated tumors, including neuroendocrine carcinomas, were developed and are discussed. The uniform application of defined terminology is essential for correlating results between different laboratories and models. It is recommended that investigators use the Bar Harbor Classification system when characterizing new GEM models or when conducting experimental interventions that may alter the phenotype or natural history of lesion progression in existing models.